GPU Modeling of Ship Operations in Pack Ice

Abstract

The paper explores the use of an event-mechanics approach to assess vessel performance in pack ice. The methodology is developed using massively parallel programming strategies on a GPU enabled workstation. A set of simulation domains, each containing hundreds of discrete and interacting ice floes is modeled. A simple vessel is modeled as it navigates through the domains. Each ship-ice collision is modeled, as is every ice-ice contact. Time histories of resistance, speed and position are presented along with the parametric sensitivities. The results are compared to published data from analytical, numerical and scale model tests. The work is part of a large research project at Memorial University called STePS2 (Sustainable Technology for Polar Ships and Structures).